README.signal revision 303975
1# $Id: README.signal,v 10.1 1995/06/23 10:28:17 bostic Exp $ 2 3There are six (normally) asynchronous actions about which vi cares: 4SIGHUP, SIGINT, SIGQUIT, SIGTERM, SIGTSTP and SIGWINCH. 5 6The assumptions: 7 1: The DB routines are not reentrant. 8 2: The curses routines may not be reentrant. 9 3: Neither DB nor curses will restart system calls. 10 11XXX 12Note, most C library functions don't restart system calls. So, we should 13*probably* start blocking around any imported function that we don't know 14doesn't make a system call. This is going to be a genuine annoyance... 15 16SIGHUP, SIGTERM 17 Used for file recovery. The DB routines can't be reentered, nor 18 can they handle interrupted system calls, so the vi routines that 19 call DB block signals. This means that DB routines could be 20 called at interrupt time, if necessary. 21 22SIGQUIT 23 Disabled by the signal initialization routines. Historically, ^\ 24 switched vi into ex mode, and we continue that practice. 25 26SIGWINCH: 27 The interrupt routine sets a global bit which is checked by the 28 key-read routine, so there are no reentrancy issues. This means 29 that the screen will not resize until vi runs out of keys, but 30 that doesn't seem like a problem. 31 32SIGINT and SIGTSTP are a much more difficult issue to resolve. Vi has 33to permit the user to interrupt long-running operations. Generally, a 34search, substitution or read/write is done on a large file, or, the user 35creates a key mapping with an infinite loop. This problem will become 36worse as more complex semantics are added to vi, especially things like 37making it a pure text widget. There are four major solutions on the table, 38each of which have minor permutations. 39 401: Run in raw mode. 41 42 The up side is that there's no asynchronous behavior to worry about, 43 and obviously no reentrancy problems. The down side is that it's easy 44 to misinterpret characters (e.g. :w big_file^Mi^V^C is going to look 45 like an interrupt) and it's easy to get into places where we won't see 46 interrupt characters (e.g. ":map a ixx^[hxxaXXX" infinitely loops in 47 historic implementations of vi). Periodically reading the terminal 48 input buffer might solve the latter problem, but it's not going to be 49 pretty. 50 51 Also, we're going to be checking for ^C's and ^Z's both, all over 52 the place -- I hate to litter the source code with that. For example, 53 the historic version of vi didn't permit you to suspend the screen if 54 you were on the colon command line. This isn't right. ^Z isn't a vi 55 command, it's a terminal event. (Dammit.) 56 572: Run in cbreak mode. There are two problems in this area. First, the 58 current curses implementations (both System V and Berkeley) don't give 59 you clean cbreak modes. For example, the IEXTEN bit is left on, turning 60 on DISCARD and LNEXT. To clarify, what vi WANTS is 8-bit clean, with 61 the exception that flow control and signals are turned on, and curses 62 cbreak mode doesn't give you this. 63 64 We can either set raw mode and twiddle the tty, or cbreak mode and 65 twiddle the tty. I chose to use raw mode, on the grounds that raw 66 mode is better defined and I'm less likely to be surprised by a curses 67 implementation down the road. The twiddling consists of setting ISIG, 68 IXON/IXOFF, and disabling some of the interrupt characters (see the 69 comments in cl_init.c). This is all found in historic System V (SVID 70 3) and POSIX 1003.1-1992, so it should be fairly portable. 71 72 The second problem is that vi permits you to enter literal signal 73 characters, e.g. ^V^C. There are two possible solutions. First, you 74 can turn off signals when you get a ^V, but that means that a network 75 packet containing ^V and ^C will lose, since the ^C may take effect 76 before vi reads the ^V. (This is particularly problematic if you're 77 talking over a protocol that recognizes signals locally and sends OOB 78 packets when it sees them.) Second, you can turn the ^C into a literal 79 character in vi, but that means that there's a race between entering 80 ^V<character>^C, i.e. the sequence may end up being ^V^C<character>. 81 Also, the second solution doesn't work for flow control characters, as 82 they aren't delivered to the program as signals. 83 84 Generally, this is what historic vi did. (It didn't have the curses 85 problems because it didn't use curses.) It entered signals following 86 ^V characters into the input stream, (which is why there's no way to 87 enter a literal flow control character). 88 893: Run in mostly raw mode; turn signals on when doing an operation the 90 user might want to interrupt, but leave them off most of the time. 91 92 This works well for things like file reads and writes. This doesn't 93 work well for trying to detect infinite maps. The problem is that 94 you can write the code so that you don't have to turn on interrupts 95 per keystroke, but the code isn't pretty and it's hard to make sure 96 that an optimization doesn't cover up an infinite loop. This also 97 requires interaction or state between the vi parser and the key 98 reading routines, as an infinite loop may still be returning keys 99 to the parser. 100 101 Also, if the user inserts an interrupt into the tty queue while the 102 interrupts are turned off, the key won't be treated as an interrupt, 103 and requiring the user to pound the keyboard to catch an interrupt 104 window is nasty. 105 1064: Run in mostly raw mode, leaving signals on all of the time. Done 107 by setting raw mode, and twiddling the tty's termios ISIG bit. 108 109 This works well for the interrupt cases, because the code only has 110 to check to see if the interrupt flag has been set, and can otherwise 111 ignore signals. It's also less likely that we'll miss a case, and we 112 don't have to worry about synchronizing between the vi parser and the 113 key read routines. 114 115 The down side is that we have to turn signals off if the user wants 116 to enter a literal character (e.g. ^V^C). If the user enters the 117 combination fast enough, or as part of a single network packet, 118 the text input routines will treat it as a signal instead of as a 119 literal character. To some extent, we have this problem already, 120 since we turn off flow control so that the user can enter literal 121 XON/XOFF characters. 122 123 This is probably the easiest to code, and provides the smoothest 124 programming interface. 125 126There are a couple of other problems to consider. 127 128First, System V's curses doesn't handle SIGTSTP correctly. If you use the 129newterm() interface, the TSTP signal will leave you in raw mode, and the 130final endwin() will leave you in the correct shell mode. If you use the 131initscr() interface, the TSTP signal will return you to the correct shell 132mode, but the final endwin() will leave you in raw mode. There you have 133it: proof that drug testing is not making any significant headway in the 134computer industry. The 4BSD curses is deficient in that it does not have 135an interface to the terminal keypad. So, regardless, we have to do our 136own SIGTSTP handling. 137 138The problem with this is that if we do our own SIGTSTP handling, in either 139models #3 or #4, we're going to have to call curses routines at interrupt 140time, which means that we might be reentering curses, which is something we 141don't want to do. 142 143Second, SIGTSTP has its own little problems. It's broadcast to the entire 144process group, not sent to a single process. The scenario goes something 145like this: the shell execs the mail program, which execs vi. The user hits 146^Z, and all three programs get the signal, in some random order. The mail 147program goes to sleep immediately (since it probably didn't have a SIGTSTP 148handler in place). The shell gets a SIGCHLD, does a wait, and finds out 149that the only child in its foreground process group (of which it's aware) 150is asleep. It then optionally resets the terminal (because the modes aren't 151how it left them), and starts prompting the user for input. The problem is 152that somewhere in the middle of all of this, vi is resetting the terminal, 153and getting ready to send a SIGTSTP to the process group in order to put 154itself to sleep. There's a solution to all of this: when vi starts, it puts 155itself into its own process group, and then only it (and possible child 156processes) receive the SIGTSTP. This permits it to clean up the terminal 157and switch back to the original process group, where it sends that process 158group a SIGTSTP, putting everyone to sleep and waking the shell. 159 160Third, handing SIGTSTP asynchronously is further complicated by the child 161processes vi may fork off. If vi calls ex, ex resets the terminal and 162starts running some filter, and SIGTSTP stops them both, vi has to know 163when it restarts that it can't repaint the screen until ex's child has 164finished running. This is solveable, but it's annoying. 165 166Well, somebody had to make a decision, and this is the way it's going to be 167(unless I get talked out of it). SIGINT is handled asynchronously, so 168that we can pretty much guarantee that the user can interrupt any operation 169at any time. SIGTSTP is handled synchronously, so that we don't have to 170reenter curses and so that we don't have to play the process group games. 171^Z is recognized in the standard text input and command modes. (^Z should 172also be recognized during operations that may potentially take a long time. 173The simplest solution is probably to twiddle the tty, install a handler for 174SIGTSTP, and then restore normal tty modes when the operation is complete.) 175